Small molecules enhance autophagy and reduce toxicity in Huntington���s disease models Sovan Sarkar1,6, Ethan O Perlstein2,3,6, Sara Imarisio4, Sandra Pineau1, Axelle Cordenier4, Rebecca L Maglathlin3, John A Webster3, Timothy A Lewis3, Cahir J O���Kane4, Stuart L Schreiber3,5 & David C Rubinsztein1 The target of rapamycin proteins regulate various cellular processes including autophagy1, which may play a protective role in certain neurodegenerative and infectious diseases2. Here we show that a primary small-molecule screen in yeast yields novel small-molecule modulators of mammalian autophagy. We first identified new small-molecule enhancers (SMER) and inhibitors (SMIR) of the cytostatic effects of rapamycin in Saccharomyces cerevisiae. Three SMERs induced autophagy independently of rapamycin in mammalian cells, enhancing the clearance of autophagy substrates such as mutant huntingtin and A53T a-synuclein, which are associated with Huntington���s disease and familial Parkinson���s disease, respectively3���5. These SMERs, which seem to act either independently or downstream of the target of rapamycin, attenuated mutant huntingtin-fragment toxicity in Huntington���s disease cell and Drosophila melanogaster models, which suggests therapeutic potential. We also screened structural analogs of these SMERs and identified additional candidate drugs that enhanced autophagy substrate clearance. Thus, we have demonstrated proof of principle for a new approach for discovery of small-molecule modulators of mammalian autophagy. The autophagy/lysosome and ubiquitin/proteasome pathways are the two main routes for protein clearance in eukaryotic cells. Proteasomes predominantly degrade short-lived nuclear and cytosolic proteins, which need to be unfolded to pass through the narrow pore of the proteasome barrel, thereby precluding clearance of large membrane proteins and protein complexes (including oligomers and aggregates). Mammalian lysosomes, on the other hand, can degrade substrates such as protein complexes and organelles. The bulk degradation of cytoplasmic proteins and organelles is largely mediated by macro- autophagy, generally referred to as autophagy1. It involves the forma- tion of double-membrane structures called autophagosomes or autophagic vacuoles, which fuse with lysosomes to form autolyso- somes (also called autophagolysosomes), where their contents are degraded by acidic lysosomal hydrolases. Autophagosomes are gener- ated by elongation of small membrane structures whose precise origins have yet to be elucidated1. Autophagy can be induced under physio- logical stress conditions such as starvation. Several protein kinases regulate autophagy, the best characterized being the mammalian target of rapamycin (mTOR), which negatively regulates the pathway in organisms from yeast to humans1. However, the targets of mTOR- dependent and mTOR-independent signaling in the autophagy appa- ratus are not well understood in mammalian systems. Recently, we described an mTOR-independent pathway in which autophagy is induced by agents that lower inositol (1) or inositol-1,4,5-tripho- sphate (IP3) (2) concentrations6. Autophagy is an important process in a variety of human diseases caused by toxic, aggregate-prone, intracytosolic proteins, which become inaccessible to the proteasome when they oligomerise2���5. These include Huntington���s disease (HD), an autosomal-dominant neurodegenerative disorder caused by a CAG trinucleotide repeat expansion (435 repeats) that encodes an abnormally long poly- glutamine (polyQ) tract in the N terminus of the huntingtin protein7. Mutant huntingtin toxicity is thought to be exposed after it is cleaved to form N-terminal fragments comprising the first 100���150 residues with the expanded polyQ tract, which are also the toxic species found in aggregates. Thus, HD pathogenesis is frequently modeled with exon 1 fragments containing expanded polyQ repeats that cause aggregate formation and toxicity in cell models and in vivo7. In addition to mutant huntingtin, autophagy also regulates the clearance of other aggregate-prone disease-causing proteins, such as those causing spinocerebellar ataxia type 3, forms of tau (which cause fronto-temporal dementias) and the A53T and A30P a-synuclein mutants (which cause familial Parkinson���s disease (PD))3���6,8,9. Auto- phagy induction reduces mutant huntingtin levels and protects against its toxicity in cell, D. melanogaster and mouse models3,4. Similar effects are seen with other polyQ-containing proteins and tau in cells and flies9. Certain bacterial and viral infections may also be treatable by autophagy upregulation, because the pathogens can be engulfed by autophagosomes and transferred to lysosomes for degradation. These Received 14 February accepted 16 April published online 7 May 2007 doi:10.1038/nchembio883 1Department of Medical Genetics, University of Cambridge, Cambridge Institute for Medical Research, Addenbrooke���s Hospital, Hills Road, Cambridge CB2 2XY, UK. 2Department of Molecular and Cellular Biology, Harvard University, 7 Divinity Avenue, Cambridge, Massachusetts 02138, USA. 3Howard Hughes Medical Institute, the Broad Institute of Harvard and MIT, 7 Cambridge Center, Cambridge, Massachusetts 02142, USA. 4Department of Genetics, University of Cambridge, Cambridge CB2 3EH, UK. 5Department of Chemistry and Chemical Biology, Harvard University, 12 Oxford Street, Cambridge, Massachusetts 02138, USA. 6These authors contributed equally to this work. Correspondence should be addressed to S.L.S. (stuart_schreiber@harvard.edu) or D.C.R. (dcr1000@cam.ac.uk). NATURE CHEMICAL BIOLOGY VOLUME 3 NUMBER 6 JUNE 2007 331 LETTERS

from the medium5. If the transgene expression level is measured at 24 h after switching off expression after an initial induction period of 48 h, one can assess whether specific agents alter the clearance of the transgene product, as the amount of transgene product decays when synthesis is stopped5. Notably, we observed that 13 SMIRs (14���18, 19a, 19b, 22, 23, 29a, 29b, 30 and 31) slowed the clearance of this autophagy substrate, whereas 4 SMERs (10, 16, 18 and 28) enhanced its clearance (Supplementary Fig. 2a���d online numbering of SMERs and SMIRs here refers to the experimental numbering scheme based on the order in which the compounds were discovered). We categorized SMIRs as possible inhibitors of autophagy if they increased A53T a-synuclein levels by 40���50% or more. Likewise, we categorized SMERs as possible enhancers of autophagy if they reduced A53T a-synuclein levels to 50% or lower than the control (Supplementary Fig. 2a���d). We then concentrated on the autophagy-inducing SMERs, which may have the greatest immediate utility as drug leads for neurodegenerative diseases (Fig. 1a). We confirmed that SMERs 10 (27), 18 (30) and 28 (36) substantially enhance A53T a-synuclein clearance in PC12 cells this effect was independent of rapamycin treatment, and efficacy was observed with the doses used in the yeast screen and also at lower concentrations (Fig. 1b). In all subsequent experiments with these compounds alone, we used the concentrations used in the yeast screen. We next studied the effect of these SMERs on another autophagy substrate: mutant huntingtin3,4. SMERs 10, 18 and 28 reduced aggregation and cell death caused by enhanced green fluorescent protein (EGFP)-tagged huntingtin exon 1 with 74 polyQ repeats (EGFP-HDQ74) in COS-7 cells (Fig. 1c). We excluded SMER 16 (subsequently redesignated SMIR 33 because additional retesting revealed it to be a suppressor of the cytostatic effects of rapamycin) from our subsequent experiments as it was toxic in COS-7 and other cell lines at the concentration that enhanced the clearance of A53T a-synuclein in PC12 cells (data not shown). No overt toxicity was observed with SMERs 10, 18 and 28 (see Supplementary Methods). We confirmed that this reduction of EGFP-HDQ74 aggregation occurs via autophagy using autophagy-competent mouse embryonic fibroblasts (MEFs) (Atg5+/+) or matched MEFs lacking the essential autophagy gene Atg5 (Atg5���/���)15. EGFP-HDQ74 aggregation was significantly (P o 0.0001) increased in untreated Atg5���/��� (auto- phagy-deficient) cells compared with untreated Atg5+/+ cells, as mutant huntingtin is an autophagy substrate (Supplementary Fig. 3a online). SMERs 10, 18 and 28 significantly reduced EGFP-HDQ74 aggregation in Atg5+/+ cells, but not in Atg5���/��� cells (Fig. 1d). Thus, these SMERs can only reduce mutant huntingtin aggregation in autophagy-com- petent cells. Indeed, these SMERs decreased EGFP-HDQ74 aggrega- tion in cells constitutively treated with the proteasome inhibitor 0 100 200 300 400 500 600 700 800 Actin EGFP-LC3-I Baf EGFP-LC3-II *** *** *** a c * ** * * 10 18 28 SMER 10 18 28 SMER Baf Baf 0 10 20 30 40 50 60 70 SMER10 SMER18 SMER28 Rap *** *** *** *** Percentage of EGFP-positive COS-7 cells with 5 EGFP-LC3 vesicles b Control SMER10 SMER28 Stable HeLa cells expressing EGFP-LC3 SMER18 20 ��m Control Baf Control Control Levels of EGFP-LC3-II (%) Figure 2 SMERs 10, 18 and 28 induce autophagy in mammalian cells. (a) COS-7 cells transfected with EGFP-LC3 construct for 4 h were treated with DMSO (control), 0.2 mM rapamycin (positive control), 47 mM SMER10, 43 mM SMER18 or 47 mM SMER28 for 16 h, and analyzed by fluorescence microscopy. The effects of treatment on the percentage of EGFP-positive cells with 45 EGFP-LC3 vesicles are shown. Error bars denote s.e.m. P o 0.0001 (all SMERs). (b) HeLa cells stably expressing EGFP-LC3 were treated with DMSO (control), 47 mM SMER10, 43 mM SMER18 or 47 mM SMER28 for 24 h. Confocal sections show cells containing EGFP-positive autophagic vesicles. Nuclei are stained with DAPI. Bar, 20 mM. (c) HeLa cells stably expressing EGFP-LC3 were treated for 4 h with DMSO (control) or 200 nM bafilomycin A1 (Baf), or with 200 nM bafilomycin A1 and 47 mM SMER10, 43 mM SMER18 or 47 mM SMER28. Cells were left untreated or pretreated with SMERs for 24 h before adding bafilomycin A1. Levels of EGFP-LC3-II were determined by immunoblotting with antibody against EGFP (above) and densitometry analysis relative to actin (below). Error bars denote s.e.m. P �� 0.0259 (baf), P o 0.0001 (SMER10), P �� 0.0003 (SMER18 and SMER28) versus control P �� 0.0025 (SMER10), P �� 0.0218 (SMER18), P �� 0.0195 (SMER28) versus bafilomycin A1. ***P o 0.001 **P o 0.01 *P o 0.05. LETTERS NATURE CHEMICAL BIOLOGY VOLUME 3 NUMBER 6 JUNE 2007 333